Drill and Learn: A Decision Making Workflow to Quantify Value of Learning

Abstract

The goal of reservoir management is to make decisions with the objective of maximizing the value creation from the oil or gas production. In order to do this models that preserve geological realism and have predictive capabilities are being developed and used. These models are commonly calibrated by using Assisted History Matching (AHM) methods which, in general, will lead to reduced uncertainty in the predicted production. Although uncertainty assessment and reduction is often an element of high-quality decision making, it is not, in itself, value-creating. Value can only be created through decisions and any decision changes resulting from AHM should be modeled explicitly. Recently there has been a spurt in the application and understanding of Value of Information workflows for reservoir management. In this talk we present a comparison of existing workflows and point out the differences between them. Following this we introduce, a practically driven approach, referred to as Drill and Learn, with elements and concepts from existing workflows to quantify the Value of Learning. The difference and definitions of Value of Information and Value of Learning (VoL) are also presented. Ensemble methods (ES-MDA and StoSAG) are used for the history matching and optimization. The results presented are obtained by applying the proposed Drill and Learn workflow on a realistic synthetic case. Sensitivities to the amount of information obtained before a closed loop exercise is performed are also investigated. We show the benefit of performing the closed loop approach to quantify the value of learning (VoL) to modify field development decisions which leads to a mature robust decision making framework.

Fast robust optimization using bias correction applied to the mean model

“…In Eq. 19, σ 2 α is an estimate of the variance of α over the domain of interest, σ 2 β is the variance of β, and n eff is the effective sample size for the observations of β (18). The regularized estimate,α r (x i ), is obtained by solving ∇ α S = 0, obtaining:…”

Closed-loop field development with multipoint geostatistics and statistical performance assessment

“…As discussed in Introduction, recent work that applied modified versions of CLFD (Morosov and Schiozer, 2016;Hanea et al, 2017), reported that application of CLFD, in some cases, has resulted in slight reduction of true-model NPV. It may seem counter-intuitive that assimilating new data and re-optimizing decision variables over the remaining span of project life, in cases has lead to a decrease in true-model NPV (which is the ultimate outcome).…”

“…The use of multiple true-models in CLFD has also been considered by Hanea et al (2017) and Shirangi and Durlofsky (2015). Hanea et al (2017), however, did not present a procedure for performance assessment (as TruMAP) and used too few cases.…”

“…The use of multiple true-models in CLFD has also been considered by Hanea et al (2017) and Shirangi and Durlofsky (2015). Hanea et al (2017), however, did not present a procedure for performance assessment (as TruMAP) and used too few cases. Barros et al (2016) also applied closed-loop optimization for multiple true-model cases, though in a different context, i.e., quantifying value of information in CLRM.…”

“…The CLFD framework has also been applied in recent work by other researchers. Morosov and Schiozer (2016) applied the CLFD to a realistic example, while Hanea et al (2017) applied the CLFD framework in simpler settings where only drilling sequence is optimized. Compared with the original work (Shirangi and Durlofsky, 2015;Shirangi, 2013), they used different reservoir modeling and different model calibration and optimization approaches in their CLFD application.…”